Slashdot Mirror
Feds Unveil Rule Requiring Cars To 'Talk' To Each Other (thehill.com)
An anonymous reader quotes a report from The Hill: The Obama administration released a long-awaited rule on Tuesday requiring all new vehicles to have communication technology that allows them to "talk" to each another, which officials say could prevent tens of thousands of crashes each year. The proposal calls for all new light-duty cars and trucks to eventually be equipped with vehicle-to-vehicle (V2V) technology, a safety system that enables cars to send wireless signals to each other, anticipate each other's moves and thus avoid crashes. The rule would require 100 percent of new vehicle fleets to have V2V technology within four years of the final rule's enactment. The proposal will be open for public comment for 90 days. The connected vehicle rule builds on previous work by the outgoing administration to accelerate the deployment of innovative safety technology. The Department of Transportation released the first-ever federal guidelines for driverless cars in September. "We are carrying the ball as far as we can to realize the potential of transportation technology to save lives," said Transportation Secretary Anthony Foxx. "This long-promised V2V rule is the next step in that progression. Once deployed, V2V will provide 360-degree situational awareness on the road and will help us enhance vehicle safety." Officials say V2V has the potential to mitigate 80 percent of non-impaired crashes and can interact with other crash avoidance systems, like automatic braking. V2V uses dedicated short-range radio communications to exchange messages about a car's speed, direction and location. The system uses that information from other vehicles to identify potentials risks and warn its driver. A pair of Democratic senators called on the agency to ensure that vehicles have "robust" cybersecurity and privacy protections in place before automakers deploy V2V.
This is going to be abused. People will have so much fun with this, it'll be unreal. Imagine a little box you can buy/build that spoofs a vehicle system and tricks all the cars in a 100 meter radius into executing an emergency stop...
So much for privacy - your car will be beaming "Hey, this guy is going 2 mph over the speed limit!" to every cop in 20 miles.
Bow down to your masters in Washington DC!
will be over. More than 20 million were killed in car accidents in the 21st century alone.
I am extremely worried about this.
1) It will be abused. Period. You know it will contain the VIN or other unique ID. So readers on the side of the road will be monitoring everyone everywhere- where you go, how fast you were going, etc. Endless tickets in the mail.
2) It will be hacked. Period. And once it is, it could cause chaos and devastation on the roads- causing other vehicles to panic and brake, others to swerve, etc. It would be one thing if this data were read-only, but we all know it will be linked into active controls. Road rage weapon. Stupid teenager prank. Whatever.
3) It will be hijacked. With active controls tieins, police cars could use spoofed info as one way to kinda remotely-control other cars. And, of course, if they can do it, so can criminals. It will give a new meaning to the word "carjacking"....
4) It will often be non-upgradable. Car manufacturers have a proven dismal track record on keeping ANYTHING updated on their cars. Once it is sold, they couldn't care less about the vehicle, unless they can somehow turn it into an endless stream of revenue.
Like any technology, there are good things and bad things with each "improvement".
Exactly... this is the part that worries me... they talk about 128 bit encryption and all that jazz, but this isn't a negotiated connection people... it's transmitting your telemetry in the blind, hoping that others will act on it. As such, everyone will be using the same encryption key, which will make it trivial for someone to transmit false information. There are literally dozens of ways I can think to abuse this capability for fun and profit.
The other issue is this: The expected range these operate at is defined by the size and quality of the antennas they intend to use, but with improved listening capability the range is much further. They claim to not transmit any specific identifying information, but if it broadcasts 10x/second, then it's pretty trivial to follow if you can receive real time. Imagine how easily you can tail a car now that you can stay out of visual range and still know exactly where they are? Tell me the police won't want that capability.
Which has more power: the hammer, or the anvil?
It IS a good thing. Not the unique ID part which is a bit scary. But cars communicating their intentions to each other. It should make roads considerably safer. HOWEVER, I suspect that implementing it is not as easy as it sounds.
It's the corner of 1st and Main in most any big US city. Or its foreign equivalent. And there are maybe 200 cars, buses, trucks, carts, mopeds, bicycles, horsecarts, etc, etc,etc in the area. And they are all squawking endlessly something like "Hi, I'm a mostly blue 2031 VW Hedgehog at latitude xx.xxxx,longitude yyy.yyyy proceeding NorthEast toward an intersection where I plan to turn left. If you need to talk to me, my friends call me $%34XQC1" And the vehicles are all using the same RF frequency band. And there are three other blue 2031 Hedgehogs in the area. And two 2032s which are visually identical except for the shape of the passenger side mirror. And, thanks to multipath GPS reception, most of the vehicles are a bit uncertain about the last three digits of their coordinates. And half those who aren't uncertain should be. And maybe it's raining. Or snowing. And there's an ambulance or fire engine that's trying to get vehicles out of its way.
Just exactly how does this nifty vehicle communication scheme sort all that out?
You can't see ANYTHING from a car, You've got to get out of the goddamned contraption and walk...Edward Abbey
Two words is no where near enough in this case. The whole idea is icredibly dangerously insane stupid, it's like what the fuck, Which shit head lobbyist managed to get this idea pushed through so their pet caimpagn contributor could believe they would make billions upon billions of dollars, fucking morons. This system is one hundred percent hackable right from the get go, seriously, it's design function is inherently hackable, it is designed to be hacked, at it's core. How, so fucking easy, just put those communications devices in all sorts of places with set transmissions, indicate of false current state for imaginary vehicle to induce dangerous evasive action in other vehicle which will trigger a real accident based up avoided the greater imaginary risk and there is nothing what so ever you can do to stop it. Absolutely fucking nuts.
Chaos - everything, everywhere, everywhen
1. Triangulation, self-identification, and multiple sensors.
2. Multiple sensors.
There's a common thread here.
Any V2V system is advice-only. My car can advise other cars that I'm in this particular place, and I'll be turning right up ahead. That lets the car to my left know that I will be getting out of its way, and it will have an opportunity to move over. Before actually doing so, it will verify that the lane is clear, that there are no other non-communicative cars, and that my car indeed slowed down to make the turn. By having multiple directional antennas on the receiving vehicle, it can tell precisely where the transmitter is, just as well as a driver looking out of the window.
The V2V system is not really intended to be treated as a paragon of truth. It can assist, but it's really just another input to another car-driving system, whether automated or human. To a human, the V2V information can be projected on the windshield as a HUD, for example, marking expected paths for nearby communicative cars. For an automated system, the V2V information can be used to plan routes with more certainty than prediction algorithms plotting likely paths based on traffic and behavior alone.
You do not have a moral or legal right to do absolutely anything you want.
Your colorfully-worded argument has convinced me that you have absolutely no idea what you're talking about.
Essentially, you argument is that the system can be used to tell vehicles false information. However, that's entirely possible today with the existing information systems. You can turn your blinker on without turning. You can randomly apply the brakes and slow down. You can weave through traffic, and disobey lane markings.
No, V2V is not going to replace common sense, or the algorithms in self-driving vehicles. If a car announces a planned path, it's not going to be trusted outright. Rather, it will be corroborated with reports from the receiving vehicle's sensor suite, other vehicles, and even fixed landmarks.
To illustrate, let's consider some example scenarios, where I will play the bad guy, with a significant technical ability, no morality whatsoever, and a strong desire to keep noisy traffic off of my quiet little street. For ease, we'll assume no functioning cryptography, focusing only on algorithmic security.
It would be pretty easy for me to make a little beacon to announce that my road's under construction. However, with no construction cones nearby, no proper roadsigns, and no road crew, vehicles would have little reason to trust my beacon. They might put a little more weight into their road sensors looking for potholes, but that's about the extent of my influence. Of course, I could post some fake construction signs, but that's illegal under current law, and the only difference is that I would be adding a radio announcing to the police that I've put up fake signs.
Let's get a little more malicious. I could spoof two vehicle transmitters, announcing an accident. Surely, traffic will route around it for a while, but eventually something will come down my road anyway, and notice that my transmitters' announced locations don't actually correspond to vehicles. In fact, it sees the road is clear, and announces that observation on the network. That causes another vehicle (which would really prefer my road anyway) to try driving past, and it corroborates the report, announcing that my transmitters are liars. Their spoofed vehicle IDs, then get flagged as fraudulent, and I have police coming to my door.
If I'm going to end up with a visit from the police eventually, I can at least make it interesting, and get the FCC involved. I can just jam all V2V traffic in my area, making my road an unknown compared to the safer routes around me. That may work for a while, but it doesn't deter any traffic that would have normally gone down my road. Without V2V communication, the vehicles fall back to their normal radar and vision sensors.
Now let's suppose I get really angry, and head out on the highway to wreak havoc. I can announce that I'm moving perfectly fine, going straight ahead. I can announce that my brakes are in great condition, that I have a full sensor suite, and all my data is trustworthy. I can announce a few spoofed vehicles just out of sight that trust my data. Everything says it's safe to come close to me, and nothing will go wrong... until I slam on my brakes. Then, the car behind me slams on his brakes automatically, because it rightly knows not to absolutely trust anything coming from V2V, and it's still watching my car with forward sensors. It also announces to the network that we're stopping, so all the cars for a mile behind me brake, too, and the ones that can shift lanes will escape. Even if there is a collision, the damage will be minimized by the slower speed and rapid self-preserving coordination. The flood of messages about the delays will be weighed more heavily than my fake announcement of clear conditions, so the affected lane will be avoided, and vehicles will move away as they're able. I've disrupted a few folks' smooth sailing, but it's not catastrophic.
No, the system isn't perfect, and it's not designed to require perfection. The people actually doing V2V work are well aware of the limitations of network communications, and are designing systems to work around malicious actors. In the worst cases, it's still better than human drivers or isolated automatons, so it's still a net benefit.
You do not have a moral or legal right to do absolutely anything you want.